Heart rate extraction using neural wavelet adaptive gain control and neural pattern processing

• US 9,345,413 B2
• Filed: 08/12/2013
• Issued: 05/24/2016
• Est. Priority Date: 08/12/2012
• Status: Active Grant

• Current Assignee: David Schie, Mike Ward
• Original Assignee: David Schie, Mike Ward
• Inventors: Schie, David, Ward, Mike
• Primary Examiner(s): Koharski, Christopher D
• Assistant Examiner(s): CAREY, MICHAEL JAMES
• Application Number:
  US13/964,923
• Publication Number:
  US 20150045684A1
• Time in Patent Office: 1,016 Days
• Field of Search: 600/509, 600/515, 600/516, 600/517, 600/521
• US Class Current: 1/1
• CPC Class Codes: A61B 5/0006 ECG or EEG signals
• CPC Class Codes: A61B 5/0022 Monitoring a patient using a global network, e.g. telephone networks, internet
• CPC Class Codes: A61B 5/0245 by using sensing means generating electric signals, i.e. ECG signals
• CPC Class Codes: A61B 5/316 Modalities, i.e. specific diagnostic methods
• CPC Class Codes: A61B 5/349 Detecting specific parameters of the electrocardiograph cycle
• CPC Class Codes: A61B 5/7207 of noise induced by motion artifacts
• CPC Class Codes: A61B 5/726 using Wavelet transforms
• CPC Class Codes: A61B 5/7264 Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems neural networks per se G06N3/00; expert systems per se G06N5/00
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems

Abstract

An improved heart rate monitor is provided that can detect and distinguish a heartbeat from an otherwise contaminated system with noise components potentially larger than the signal of interest. Embodiments of the inventive monitor have an amplification system that eliminates large noise components so as not to saturate the system during detection of a desired low amplitude signal. In embodiments the elimination of noise components is accomplished through wavelet decomposition, and the removal of undesired components including interference components during adaptive gain control (AGC), in addition to hunting algorithms which minimize the error with techniques such as neural network least mean squares type back propagation algorithms.

Claims

1. A heart rate measurement device comprising:

a wavelet template;

an instrumentation amplifier;

a set of two electrodes configured to be coupled to a subject'"'"'s skin, and in electrical communication with and providing a set of signals to said instrumentation amplifier;

an adaptive gain control (AGC) algorithm;

an analog or digital filter in a feedback path of said instrumentation amplifier for at least one of removal of saturating signals and unwanted waveform components responsive to said AGC algorithm;

wherein said AGC algorithm utilizes said wavelet template and responds to the correlation of said wavelet template to a measured signal obtained from said set of signals by said instrumentation amplifier; and

wherein said device is configured to capture said set of signals utilizing a high common-mode resection ratio (CMRR) termination, which does not include a DC path to ground or overload said electrodes.

2. The device of claim 1 wherein said instrumentation amplifier further comprises:

two input amplifiers, where a level of gain is set by at least one of a variable resistor or impedance between the two input amplifiers, as well by feedback components around the two input amplifiers; and

a summing amplifier.

3. The device of claim 1 wherein said instrumentation amplifier includes a common mode driver, a common mode accelerator, and an active feedback network.

4. The device of claim 1 wherein said instrumentation amplifier is configured for baseline capture by changing a set of feedback components such that saturation recovery is accelerated.

5. The device of claim 4 wherein the baseline capture is obtained with a third electrode driven to create a baseline between said set of signals from said two electrodes placed a minimum impedance away from each other.

6. The device of claim 1 wherein said measured signal is sent wirelessly to the cloud or a remote server for digital processing.

7. The device of claim 1 wherein said instrumentation amplifier precedes a hardware bessel filter.

8. The device of claim 1 wherein said measured signal is analyzed by a wavelet decomposition that builds up a three or more dimensional image of said measured signal.

9. The device of claim 8 wherein said three or more dimensional image is analyzed with pattern recognition methods to determine heart beat components responsive to a training algorithm.

10. The device of claim 9 wherein said pattern recognition methods are a neural network trained to recognize heart beat components from said measured signal.

11. The device of claim 10 wherein a periodic loss of signal is recovered through a prediction based on said neural network, where a rate or waveform may be reconstructed from discontinuous snapshots of heart beat trains.

12. The device of claim 1 wherein said wavelet is in resonance with one or more of the P, QRS and/or T reference components of a heart beat signal.

13. The device of claim 12 wherein said wavelet is configured to meet the following:

transform and inverse transform and decomposition requirements;

admissibility requirements including meeting the Perceval theorem condition; and

convergence admissibility conditions.

14. The device of claim 1 further comprising a thresholding algorithm utilizing an error minimization algorithm to iteratively hunt for an optimal wavelet correlation to a train of heart beats.

15. The device of claim 14 further comprising an error minimization algorithm, wherein said error minimization algorithm thresholds the decomposition of a set of 2^n wavelet coefficients and one of optimizes relevant components or minimizes or removes any components of said set of signals which saturate said instrumentation amplifier and are unrelated to said heart rate signal by thresholding only relevant wavelet coefficients from said signal in a path of said digital filter;

wherein removed components comprise at least one of 60 Hz noise, muscle movement, electrode connection and misconnection, electromagnetic machine noise, respiration, motion artifacts, or other interference sources.

16. The device of claim 2 wherein said instrumentation amplifier comprises at least one of two digital or two neural network formed transfer functions replacing said two input amplifiers and feedback components in the first stage, where a gain and selection of waveform components to pass through said transfer functions are set in conformance with an AGC algorithm.

17. The device of claim 16 wherein said AGC algorithm includes a wavelet decomposition and error minimization function such that only those wavelet decomposition components responsive to a heart rate waveform are allowed through said first stage of said instrumentation amplifier.

18. A method for conditioning a measured signal of a heart rate measurement device comprising:

providing a heart rate measurement device comprising;

a wavelet template;

an instrumentation amplifier;

a set of two electrodes configured to be coupled to a subject'"'"'s skin, and in electrical communication with and providing a set of signals to said instrumentation amplifier;

an adaptive gain control (AGC) algorithm;

an analog or digital filter in a feedback path of said instrumentation amplifier for at least one of removal of saturating signals and unwanted waveform components responsive to said AGC algorithm;

wherein said AGC algorithm utilizes said wavelet template and responds to the correlation of said wavelet template to a measured signal obtained from said set of signals by said instrumentation amplifier; and

wherein said device is configured to capture said set of signals utilizing a high common-mode rejection ratio (CMRR) termination, which does not include a DC path to ground or overload said electrodes;

removing artifacts outside a heart rate window by applying one or more Bessel filters to create filtered waveform;

smoothing said filtered waveform with a smoothing function by averaging points around each point to create a baseline;

subtracting the baseline from the filtered waveform; and

performing adaptive gain control (AGC) on the filtered signal in conformance with a three or more dimensional image produced using iterative decomposition wavelet template algorithm.

19. A method of heart waveform extraction comprising:

providing a wavelet template responsive to PQRST components of said heart waveform;

providing a contaminated signal containing said heart waveform as well non-ideal components measured by set of two electrodes configured to be coupled to a subject'"'"'s skin;

using a processor to provide an at least three dimensional image created by deconstructing said contaminated signal utilizing said wavelet template by said processor; and

providing a pattern recognition neural network trained to recognize heart beat components from an image created from similar wavelet deconstructed signals including noise components, wherein said at least three dimensional image is applied to said neural network trained to produce an output vector responsive to recognized heart beats.

20. A heart rate measurement device comprising:

a wavelet;

an instrumentation amplifier;

a set of two electrodes configured to be coupled to a subject'"'"'s skin, and in electrical communication with and providing a set of signals to said instrumentation amplifier;

an adaptive gain control (AGC) algorithm;

an analog or digital filter in a feedback path of said instrumentation amplifier for at least one of removal of saturating signals and unwanted waveform components responsive to said AGC algorithm;

wherein said AGC algorithm utilizes said wavelet and responds to the correlation of said wavelet to a measured signal obtained from said set of signals by said instrumentation amplifier;

wherein said measured signal is analyzed by a wavelet decomposition that builds up a three or more dimensional image of said measured signal, said three or more dimensional image analyzed with pattern recognition methods to determine heart beat components responsive to a training algorithm; and

wherein said pattern recognition methods are a neural network trained to recognize heart beat components from said measured signal.